1

Flammability properties of British heathland and moorland 1

vegetation: models for predicting fire ignition 2

3

4

Victor M. Santanaa,b,*, Rob H. Marrsa 5

6

aSchool of Environmental Sciences, University of Liverpool, Liverpool L69 3GP, UK. 7

bFundación de la Generalitat Valenciana Centro de Estudios Ambientales del 8

Mediterráneo (CEAM), Parque Tecnológico Paterna. C/ Charles Darwin, 14, E-46980 9

Paterna, Valencia. Spain. 10

*Corresponding Author. Tel: +44 (0) 1517955172; E-mail address: vm.santana@ua.es 11

12

13

14

15

16

17

2

Abstract 18

Temperate ecosystems, for example British heathlands and moorlands, are predicted to 19

experience an increase in severe summer drought and wildfire frequency over the next 20

few decades. The development of fire ignition probability models is fundamental for 21

developing fire-danger rating systems and predicting wildfire outbreaks. This work 22

assessed the flammability properties of the fuel complex of British moorlands as a 23

function of their moisture content under laboratory conditions. Specifically, we aimed to 24

develop: (1) models of the probability of fire ignition in peat/litter fuel-beds (litter of 25

four different plant species, Sphagnum moss and peat); (2) flammability properties in 26

terms of ignitability, sustainability, consumability and combustibility of these peat/litter 27

fuel-beds; (3) the probability of ignition in a canopy-layer of Calluna vulgaris (the 28

most dominant heath/moor species in Britain) as a function of its dead-fuel proportion 29

and moisture content; (4) the efficacy of standardized smouldering and flaming ignition 30

sources in developing sustained ignitions. For this, a series of laboratory experiments 31

simulating the fuel structure of moor vegetation were performed. The flammability 32

properties in peat/litter fuel-beds were influenced strongly by the fuel moisture content. 33

There were small differences in moisture thresholds for experiencing initial flaming 34

ignitions (35-59%), however, the threshold for sustained ignitions (i.e., spreading a 35

fixed distance from the ignition point) varied across a much wider range (19-55%). 36

Litter/peat fuel-beds were classified into three groups: fuel-beds with high ignitability 37

and combustibility, fuel-beds with high levels of sustainability, and fuel-beds with low 38

levels in all flammability descriptors. The probability of ignition in the upper Calluna-39

vegetation layer was influenced by both the proportion of dead fuels and their moisture 40

content, ranging from 19% to 35% of moisture as dead fuel proportion increased. 41

Smouldering sources were more efficient in igniting peat/litter fuel-beds but in the 42

3

Calluna-vegetation layer flaming sources performed better. This work can assist in 43

improving the predictions of fire-rating systems implemented in British moorlands, by 44

providing better warnings based on critical moisture thresholds for various fuel types. 45

46

Keywords: combustibility, consumability, fire-rating systems, fuel moisture content, 47

ignitability, sustainability. 48

49

50

51

52

53

54

55

56

57

58

59

60

61

62

63

4

1. Introduction 64

Developing management strategies to face novel disturbance regimes associated with 65

climate change are fundamental for mitigating their effects (Allen et al., 2013; Marino 66

et al. 2011). Changes predicted to occur as a result of global climate-change over the 67

next few decades are that temperate ecosystems will experience an increase in severe 68

summer drought and wildfire frequency (Krawchuk et al., 2009). It is well known that 69

the occurrence of wildfires in these systems is often exacerbated under drought 70

conditions because there is no limitation in fuel availability (Pausas and Ribeiro, 2013). 71

At present, however, adaptive strategies for facing these future scenarios are in the early 72

stages of development, for example through the implementation of fuel management 73

strategies to reduce fire impacts, improved education to minimize fires started by arson 74

and the development of rating systems for forecasting fire outbreaks (Albertson et al., 75

2009; Allen et al., 2013; Davies and Legg, 2008). 76

Even though fire has played a role in shaping many temperate ecosystems, little is 77

known about the flammability properties of the component species (van Altena et al., 78

2012). Previous studies have been centered mainly in ecosystems with a high burning 79

frequency where wildfire is an ongoing problem, e.g. in Mediterranean systems. These 80

studies deal with the general ability of vegetation to burn (flammability as proposed by 81

Anderson, 1970; Martin et al., 1994); but this is usually broken down into four 82

components (1) ignitability, how easily the fuel ignites, (2) sustainability, how well the 83

combustion proceeds, (3) consumability, the amount of fuel lost during the fire, and (4) 84

combustibility, the velocity or intensity of the combustion. One major shortcoming of 85

these studies is that they have traditionally just used discrete fuel elements (e.g. leaves, 86

twigs), neglecting the possible interactions aggregated within a more complex and 87

realistic fuel-bed (Fernandes and Cruz, 2012). For instance, thin and small leaves can 88

5

ignite easily on an individual basis, but burn with difficulty when presented in litter 89

beds (Scarf and Westoby, 2006). In this respect, the flammability of the above-ground 90

vegetation is defined mainly by the structural arrangement of the fuel materials and 91

factors such as the size-distribution of the fuel elements, the dead:live ratio and bulk 92

density (Chandler et al., 1983; Santana et al., 2011), whereas small-scale intrinsic 93

properties (e.g. specific gravity, mineral content, chemical composition) have a lesser 94

effect because usually there is a low range of inter-species variation (Fernandes and 95

Cruz, 2012). Moreover, when modeling vegetation flammability, it is also necessary to 96

consider environmental conditions (e.g., moisture, temperature, wind speed and 97

direction), but especially the fuel moisture content (FMC, Marino et al., 2010; Plucinski 98

et al., 2010). All of these environmental variables interact with, and moderate, 99

flammability. 100

Heathlands and moorlands in the United Kingdom (UK) are temperate ecosystems 101

dominated mainly by the dwarf-shrub Calluna vulgaris (L.) Hull (Gimingham, 1972). 102

The vegetation fuel-complex is usually composed of three main strata: (1) the shrub 103

stratum of the above-ground vegetation, i.e. the Calluna; (2) an understory stratum of 104

litter and bryophytes; and, (3) the soil which is often an acidic podzol with a clear 105

organic mor horizon (lowland heaths) or peat (upland moors). Most heaths and moors in 106

the UK systems are originally anthropogenic, and are sustained by means of grazing and 107

burning practices that combine to prevent succession to more mature woodlands 108

(Gimingham, 1972). Land managers periodically apply rotational burning to produce a 109

mosaic of different stages of recovery and that optimizes productivity, diversity and 110

environmental services (Harris et al., 2011). The legal burning period is from October to 111

mid-April (Anon, 2007), when soils are wet and/or frozen and damage to understory 112

species and peat is minimized. However, one of the greatest threats to these ecosystems 113

6

is wildfire; these occur mainly in spring (March to April) and summer (July and 114

August) (Albertson et al., 2009). Spring wildfires comprise mainly the above-ground 115

vegetation because soils are usually still very wet, but the shrub stems are highly-116

desiccated as consequence of winter frosts (Davies and Legg, 2008). Summer wildfires, 117

in contrast, can be extraordinarily damaging because the surface peat can be dry and 118

once ignited, it can smoulder for many months (Rein et al., 2008). 119

Wildfires in British moorlands are usually caused by human negligence or malice, 120

but there is still little documented evidence about this (McMorrow, 2011). Two types of 121

ignition sources have been identified as being probably important: (1) smouldering 122

sources (e.g., such as discarded cigarettes, lost barbecues embers, hot particles dropped 123

from power lines, etc.) and (2) flaming sources (e.g. escaped prescribed burns, arson, 124

etc.) (Schmuck et al., 2012). There is, therefore, a need for a better understanding of the 125

ignition efficiency of these different sources in developing self-sustained wildfires on a 126

range of ecosystems. Moreover, the variable nature of fires (i.e., canopy fires often burn 127

independently from ground-layer fuels; Davies and Legg, 2008) means that separated 128

assessments are needed in the different strata. 129

The litter layer is the medium in which ignition is most likely to occur (Davies and 130

Legg, 2011); nonetheless the probability of ignition and subsequent fire impacts can 131

differ significantly between species because the different flammability properties of 132

their litters (Plucinski and Anderson, 2008; Scarf and Westoby, 2006). On the other 133

hand, when canopy fires occur in spring, ignition is strongly related to the moisture 134

content of dead material in the canopy fuel (Davies and Legg, 2011). Therefore, 135

estimation of moisture thresholds for fire ignition in each fuel type is of fundamental 136

relevance for predicting fire danger (Davies and Legg, 2008). Assessing these 137

7

thresholds is difficult under field experimental conditions, because dead-fuels are 138

usually inter-mixed with green fuels within the shrub layer (Davies and Legg, 2011). 139

The development of fire ignition probability models that incorporate the FMC of live 140

and dead canopy material, the peat/litter layer and peat are needed to develop improved 141

fire-rating systems for UK moorlands. The main aim of this work is, therefore, to assess 142

the flammability properties of a range of common species that could contribute to the 143

fuel complex of British heathlands/moorlands. To do this, we carried out a series of 144

laboratory experiments simulating the fuel structure of heath/moor vegetation under 145

controlled conditions. Specifically we aimed to develop: 146

(1) Predictive models of the probability of fire ignition in peat/litter fuel-beds (litter 147

of different plant species, Sphagnum moss and peat), using FMC as the 148

predicting variable. 149

(2) Flammability properties in terms of ignitability, sustainability, consumability 150

and combustibility of the different peat/litter fuel-beds by means of easily 151

measurable descriptors. 152

(3) Predictive models for the probability of ignition in Calluna-dominated 153

heathlands/moorlands as a function of its dead-fuel proportion and FMC. 154

(4) An assessment of the efficacy of standardized smouldering and flaming ignition 155

sources in developing sustained ignitions. 156

157

2. Methods 158

Plant material was collected throughout the summer and autumn of 2012 from three 159

heathlands/moorlands in: (1) North Wales (Sphagnum spp. L. and Vaccinium myrtillus 160

L.; 53°04’N, 3°10’W), (2) Peak District Natural Park (Calluna vulgaris (L.) Hull, 161

8

Empetrum nigrum L. and peat; 53°25’N, 1°10’W) and (3) Wirral (Ulex europaeus L.; 162

53°21’N, 3°10’W). Hereafter, species are referred by their generic names. The plant 163

material (stems and shoots) from the dwarf shrubs (Calluna, Vaccinium, Empetrum and 164

Ulex) were collected by cutting with secateurs near the ground surface. Surface cores 165

(0-5 cm depth) of mosses formed by Sphagnum and peat were collected by excavation. 166

The sampled material was transported in plastic bags to the laboratory, where it was 167

used to reconstruct (a) peat/litter fuel-beds and (b) stands of Calluna vegetation. 168

169

2.1. Laboratory preparation of the peat/litter fuel-beds 170

Initially, the plant material was placed in paper bags and oven-dried at 80ºC for 24 h. 171

This allowed easy separation of the leaves from the stems; the leaves were then used to 172

reconstruct pure leaf litter-beds for each plant species within a circular tray of 250 mm 173

diameter and 20 mm depth (Fig. S1a). For peat, the upper part of peat cores was cut 174

with a knife. Then, they were carefully prepared to have the same dimensions of the tray 175

used. The tray was similar to that used by Plucinsky and Anderson (2008) and was 176

constructed using a fireproof, fibre-base and sides of 0.5 mm stainless steel mesh. Filled 177

trays were weighed before and after each test to assess fuel consumption; the bulk 178

density was calculated from these weights and the known volume of the fuel-bed (982 179

cm3). 180

Ignition experiments were run with each of the litter/peat materials; in these 181

experiments the plant materials were manipulated to produce a range of fuel moisture 182

contents. To do this, fuels were placed into sealed plastic bags and moistened until they 183

reached the desired water content. The bags were then placed within an oven at 60ºC 184

and mixed twice daily for two days to produce a uniform moisture content. The FMC 185

was then determined as the percentage of dry mass before each test using gravimetric 186

9

method (taking a sub-sample from each prepared fuel-bed and oven drying at 80ºC for 187

two days). 188

Specific traits of the peat/litter fuel-beds, e.g. the surface area of the material, 189

surface-area to volume ratio, mineral content and heat of combustion were also 190

assessed. Assessments for litter fuel beds were made for leafs (the most part of fuels 191

used), avoiding shoots and stems. The surface area of the materials was assessed by 192

scanning samples of the material using an HP Scanjet 4850 (200dpi resolution) and 193

image processing software (ImageJ; http://rsbweb.nih.gov/ij/; accessed 16 August 194

2013). The area of Calluna and Ulex was calculated assuming a cylindrical shape. 195

Volume was measured by putting material in a pycnometer (van Altena et al., 2012). 196

Fuels were ashed in a muffle furnace at 550oC for 2h to assess their mineral content 197

(Frandsen, 1997), and heat of combustion was determined using a bomb calorimeter 198

(e2K, Digital Data Systems, South Africa). 199

200

2.2. Laboratory preparation of Calluna vegetation 201

Dead and live shoots were collected near the ground surface and transported to the 202

laboratory to produce simulated Calluna vegetation arrays. These arrays were 203

reconstructed using a similar structure to that used by Plucinski et al. (2010). This 204

consisted of two wire cages with 64 cells where individual shrub clippings were placed 205

upright (Fig. S1b). This structure was a 20 x 20 cm square and the area was sufficient 206

for demonstrating that ignition of fires was sustainable (Plucinski et al., 2010). Here, the 207

physical structure of Calluna vegetation was simulated using representative values from 208

Davies et al. (2009), who, in extensive work within a series of age-stages of Calluna 209

(building, late-building and mature) on British moorlands, showed that bulk density 210

ranged between 3.5-5 kg·m-3 and height varied between 15-45 cm. Therefore, we 211

10

produced simulated Calluna stands with shoots 30 cm tall with a bulk density of 4 212

kg·m-3; this was kept constant in all experimental runs. 213

The two key variables manipulated in this study because of their known influence on 214

fire ignition were: (1) the proportion of dead-fuel in the vegetation, and (2) the FMC of 215

the dead-fuel (Davies and Legg, 2011). Three levels of dead-fuel proportion (20, 40 and 216

60%) were reproduced with the aim of simulating different states of shrub maturation 217

(Davies et al., 2009; Davies and Legg, 2011). Shrub arrays were reproduced taking into 218

account the stratified structure of Calluna vegetation, with the dead-fuel accumulating 219

in the lower part of the canopy (Davis and Legg, 2011). For this, we cut live shoot 220

clippings (<5mm stem diameter) to a height of 30 cm and dead-fuels shoots to a height 221

of 15 cm. The FMC of live shoots was maintained constant at near field values by 222

maintaining their bases in water-filled buckets, but the exact value was determined as a 223

percentage of dry mass before each test (mean: 51.8 ± SD: 4.7, n=240). The amount of 224

dead-fuel was determined by drying the dead shoots at 80oC for two days and then 225

weighing them. The FMC in dead-fuels was modified in a similar way to litter fuels, by 226

enclosing in plastic bags and wetting the shoots to a desired level. The exact level of 227

moisture was assessed by putting an additional sub-sample within the plastic bag. This 228

sub-sample was separated by a permeable nylon bag that allowed the fuel have the same 229

moisture content as the fuel to be burned. Bulk density was kept constant by 230

proportionally decreasing or increasing the amount of live shoots regard to dead-fuel 231

proportion, but always taking into account their moisture content. 232

233

2.3. Ignition source 234

The effects of the two main types of wildfire ignition sources, smouldering sources and 235

flaming sources, were tested. The smouldering source was used in the litter/peat 236

11

experiments and both sources were used in the Calluna litter and vegetation 237

experiments. 238

The smouldering source was created electrically using a nichrome wire (300 mm 239

long and 0.5 mm width) connected to a power supply (Skytronic 650.682 Bench Top 0-240

30V 10A, Netherlands). The wire acted as a resistance and warmed until it became red 241

(temperatures ca. 600-700ºC, measured using a thermocouple K type). The central part 242

of the wire was shaped into a compact cylinder 6 mm long and 7 mm diameter by 243

giving 7 turns to the wire (Fig. S1c). The aim was to simulate the effect of a cigarette 244

end or a stray ember. To ignite litter trays, the wire cylinder was lowered into the central 245

part of the tray within the first cm of the fuel surface. In the Calluna vegetation arrays, 246

the wire cylinder was placed at the front side of the fuel structure at a height of 70 mm 247

within the dead-fuel. A power of 100W was supplied for 5 min in each test. 248

The flaming ignition source was provided through the use of commercial kerosene 249

ignition pills, designed for barbecues (Zip, Standard Brands, UK; Fig. S1d). The pills 250

were rectangular (19 x 17 x 12 mm; L x W x W). The flaming ignition source, when lit, 251

remained on fire for 383 ± 32 s (mean ± SD, n=6) and the flames reached a maximum 252

height of 101 ± 7 mm. 253

254

2.4. Experimental conditions 255

All ignition experiments were performed within a glasshouse with a temperature of 17.1 256

± 5.9°C (Mean ± SD, n=528) and a relative humidity of 44.8 ± 15.4%. The incidence of 257

wind in these types of experiments has an increasing effect in igniting litter beds 258

(Marino et al. 2010). In order to simulate wind, a domestic fan was used to provide a 259

constant air flow of 0.3 m·s-1 (measured with an anemometer-Viking ART 02041, 260

Sweden) in the central point of the tray. Wind speed was minimal in order to be 261

12

conservative in obtaining flammability parameters. Air-flow was supplied at angle of 262

45º to the experimental trays to avoid fuel particles being blown-off (Marino et al. 263

2010). 264

265

[INSERT TABLE 1] 266

267

2.5. Assessment of fuel flammability properties 268

The probability of ignition in litter, mosses and peat was assessed as a function of FMC. 269

In addition, this ignition was assessed in two different ways: (1) initial flaming ignition 270

and (2) sustained ignition. Initial flaming ignition was considered successful if flames 271

appeared after the ignition source was applied (only for the smouldering source). 272

Sustained ignition was considered positive if the fire front reached the tray edge. The 273

distance from the ignition point to the edge (125 mm) allowed enough fire development 274

to demonstrate sustainability of fire spread (Plucinsky and Anderson, 2008). A note was 275

made if the fire front reached the edge of the tray as well as whether the fire front was a 276

smouldering or flaming one. A minimum of 40 tests were performed for each fuel type. 277

In addition, flammability components (ignitability, sustainability and consumability, 278

combustibility) of the fuel types were determined using easily measurable descriptors 279

(Table 1). The time elapsed by the fire front to reach the edge of the tray and for the end 280

of combustion was recorded directly with a chronometer. This allowed us to estimate 281

the rate of spread (ROS) and mass loss rate (MLR). All tests were recorded with a 282

digital camera separated 50 cm horizontally from the tray, providing an estimate of time 283

to ignition (TTI), flaming time (FT) and flame height (FH, using a ruler located behind 284

the tray). The maximum temperature achieved (TMAX) and the time above 300oC 285

13

(T300) were obtained using four thermocouples (1 mm thick, K type) placed equally 286

around the tray and linked to a data logger (OM-DAQPRO-5300, Omega, USA). The 287

tip of each thermocouple was placed 6 cm from the center and at 1 cm of depth below 288

the fuel surface. Measurements were taken every second and the mean value of 289

temperatures from the four thermocouples was estimated for each sample. 290

The probability of sustained ignition was also assessed for Calluna vegetation. 291

Ignition was considered successful if fire reached the bottom part of the cage (20 cm). 292

The shrub support cages were weighed before and after fires in order to determine fuel 293

consumption. These Calluna vegetation experiments included 40 tests for each of the 294

three dead-fuel proportions and the two ignition sources (240 tests in total). 295

296

2.6. Statistical analysis 297

Differences in specific traits of peat/litter fuel-beds were analyzed by means of one-way 298

ANOVA with Bonferroni pair-wise comparisons. The probability of ignition was 299

modelled using Generalized Linear Models (GLM) with a binomial error distribution 300

and a logit-link function for each peat/litter fuel-bed (Crawley, 2012). Initially, we 301

considered FMC, air temperature and relative humidity as predictor variables. Then, 302

starting from the full model, the minimum adequate GLM was obtained by sequential 303

removal of non-significant model terms (Analysis of deviance, F tests, P>0.05; 304

Crawley, 2012). Because temperature and relative humidity were mainly constant 305

throughout the experiment, only FMC was selected as significant in all cases. The 306

goodness of fit was measured by Nagelkerke’s pseudo R2 statistic, and the area under 307

Receiver Operating Characteristic (ROC) curve used to determine the discriminative 308

ability of the models over a range of cut-off points (Hosmer and Lemeshow, 2000). 309

Thereafter, the FMC at which 50% of ignitions were successful (M50) was estimated for 310

14

each fuel type. The maximum FMC at which a successful ignition occurred (Mmax) was 311

also estimated. M50 values were obtained by using the logit model whereas Mmax values 312

were from observed data. In order to ascertain the influence of the different specific fuel 313

traits in ignition, the relationships between these specific traits and the M50 values were 314

assessed by mean of linear regressions. The different efficiency in initiating sustained 315

ignitions between smouldering and flaming ignition sources in Calluna litter was tested 316

by means of analysis of deviance. Flammability descriptors of each peat/litter fuel-bed 317

were modelled as a function of FMC using GLMs with a Poisson error distribution and 318

a log-link function. A summary of all the Minimum Adequate Models derived from the 319

GLM analysis is provided in Table S1. 320

The probability of ignition of the Calluna vegetation was also modelled using GLM 321

with a binomial error distribution and a logit-link function. Initially, we considered the 322

FMC of the dead-fuel, dead-fuel proportion, ignition source, air temperature and relative 323

humidity as predictor variables. Interactions between FMC of dead-fuel, dead-fuel 324

proportion and ignition source were also included in the initial model. In these models 325

the flaming ignition source was used as the baseline. As before, the final model was 326

obtained by sequential removal of non-significant terms (Analysis of deviance, F tests, 327

P>0.05). M50 and Mmax values for each dead-fuel proportion and ignition source were 328

also calculated as above. All statistical analyses were performed in the R statistical 329

environment (version 2.14.2., Development Core Team 2012, Vienna). 330

331

[INSERT TABLE 2] 332

333

334

15

3. Results 335

3.1. Flammability of peat/litter fuel-beds 336

There was considerable variation in the basic properties of the fuel-bed materials. In 337

terms of bulk density there were three groups (Table 2): peat had the greatest bulk 338

density at 289 kg m-3, Calluna and Empetrum litter had intermediate values (112-149 kg 339

m-3) and Sphagnum, Vaccinium and Ulex had least (<50 kg m-3). In terms of surface 340

area Empetrum and Ulex had the lowest whereas Calluna, Vaccinium and Sphagnum 341

had the greatest. Area to volume ratio was higher for Calluna and Vaccinium, 342

intermediate values were for Ulex, and Empetrum and Sphagnum were the lowest 343

(Table 2). Peat and Vaccinium had the largest high mineral content (11% and 6% 344

respectively) in comparison to the other species (2-3%). The heat of combustion was 345

greatest in the dwarf shrub, intermediate in peat and least in the Sphagnum (Table 2). 346

The probability of ignition was well explained by FMC (Table 3). When smouldering 347

ignition sources were applied, Sphagnum had the largest M50 values for both the 348

threshold of initial and sustained ignition (56.5% and 54.6% respectively). Litter of 349

Ulex also had high values with 51.4% and 34.5% (Table 3). In contrast, litter of 350

Calluna, Empetrum and Vaccinium had high values for the thresholds of initial ignition 351

(53.6%, 59.2% and 46.8%), but the threshold of sustained ignition was very much lower 352

(26.9%, 19.1% and 25.1%). Peat had low values for both variables (34.9% and 21.6%; 353

Table 3). Mmax values followed similar trends for Calluna, Empetrum, Vaccinium and 354

Ulex, with an increase of ca. 5-15% with respect to M50 values. In contrast, Mmax values 355

for Sphagnum and peat experienced an increase of ca. 25%. The threshold of sustained 356

ignition decreased when a flaming source was applied, as for example, observed in 357

Calluna litter (Analysis of Deviance, F= 36.65, P<0.001), where M50 values decreased 358

from 26.9% to 15.2% (Table 3B). No clear relationships were found between specific 359

16

fuel traits and M50 values, either for initial and sustained ignition (Figure 2S). Only the 360

mineral content of fuels had a significant relationship for the initial ignition (R2=0.851, 361

P=0.009). 362

363

[INSERT TABLE 3] 364

365

Sustained ignitions which spread successfully to the edge of the tray occurred mainly 366

as smouldering fires. Successful sustained ignitions as a flaming fire occurred with 367

Sphagnum and Ulex, and only when the FMC was under ca. 30%. 368

Flammability descriptors were clearly influenced by FMC (Fig. 1). The litter/peat 369

materials could be classified into three groups on the basis of these relationships. Group 370

1 comprised Ulex and Sphagnum; these species experienced the highest levels of 371

ignitability (low TTI and high ROS), consumability (high MLR and RMF) and 372

combustibility (high FH). Group 2, composed of Calluna, Empetrum and Peat, 373

experienced lower values of these flammability descriptors, but had the highest 374

sustainability values (high FT and T300). Finally, Vaccinium experienced low values 375

for all flammability descriptors. No large differences were observed in TMAX between 376

any of the fuel-beds; although Empetrum and Vaccinium experienced slightly lower 377

TMAX values (Fig. 1). 378

379

[INSERT FIGURE 1] 380

381

382

17

3.2. Flammability of Calluna vegetation 383

The selected GLM for the probability of ignition included the dead-fuel moisture, dead-384

fuel proportion, ignition source and the interaction between the ignition source and 385

dead-fuel proportion as predictor variables (Table 4). For flaming ignition sources, 386

FMC was the main factor controlling the probability of ignition; the proportion of dead-387

fuel did not affect it significantly. M50 values were all around 30% of FMC and Mmax 388

around 45-50%. In contrast, for the smouldering ignition source, the proportion of dead-389

fuel increased the ignition threshold with M50 values increasing from 19% to 35% of 390

FMC as the dead fuel proportion increased from 20% FMC to 60% (Fig. 2). In general, 391

Mmax values followed a similar trend compared to M50 values, with an increase of ca. 5-392

10%. 393

394

[INSERT TABLE 4 AND FIGURE 2] 395

396

4. Discussion 397

4.1. Fire danger in peat/litter fuel-beds 398

In British heathlands and moorlands, most wildfires have been shown to start within the 399

litter layer (Davies and Legg, 2011), from where it can spread upwards into the canopy 400

and downwards into the underlying peat (Plucinski et al., 2010; Rein et al., 2008). 401

Initially, flaming ignition in the litter fuel-beds can propagate fire to the upper canopy 402

through contact with the lower branches of vegetation. Here, we suggest that there are 403

small differences in the moisture threshold for initial flaming combustion between litter 404

fuel-beds of the different species (M50 from 47-59%). Nonetheless, despite these low 405

differences in initial ignition probability, there is variation in other flammability 406

18

properties that may confer different efficiencies in fire propagation. Fuel-beds with high 407

ignitability (low time-to-ignition, TTI) and combustibility (high flame height, FH), e.g. 408

Ulex and Sphagnum, may make contact quickly with higher branches in the vertical 409

structure of vegetation and expedite flame upwards transfer. In contrast, fuel-beds with 410

high sustainability (high flaming time, FT), e.g. Calluna and Empetrum, may maintain a 411

flame for longer and hence could propagate fires more easily because the flame will 412

have a longer contact time. In this sense, further studies assessing which flammability 413

properties (ignitability and combustibility vs. sustainability) are more important in 414

propagating fire to the aboveground vegetation are needed. 415

In contrast to the thresholds of producing initial flaming ignition, the thresholds of 416

sustained ignitions within the different litter fuel-beds varied across a wide range of 417

FMC (M50 from 19-55%). Fuel-beds able to keep sustained ignitions at higher FMC 418

values were again Ulex and Sphagnum. This ability was probably a consequence of their 419

high consumability and combustibility, observed in their high values of ROS, MLR and 420

RMF. In addition, these fuel-beds were the only ones able to spread as a flame, albeit at 421

low FMC values. The other fuel-beds, Calluna, Empetrum and Vaccinium, experienced 422

the opposite trends. No clear relationship was found among fuel-bed traits of the species 423

studied and flammability properties. Bulk density was probably the most influential 424

trait; low-density litter beds composed of big particles tend to pack more sparingly and 425

allow better aeration for fire development (Ganteaume et al. 2011; Plucinski and 426

Anderson 2008; Scarf and Westoby 2006). For all but one species, flammability 427

properties followed this pattern, as we found that species with the lowest bulk densities 428

(Ulex and Sphagnum) experienced a greater flammability than species with high bulk 429

densities (Calluna, Empetrum and Peat). The exception was Vaccinium which had a low 430

bulk density but its flammability was low. It is possible that this result was brought 431

19

about through interactions with other fuel traits, for example area to volume ratio, 432

mineral content, the physical arrangement of fuel particles or others factors not 433

examined here. In addition, it is worth noting that area to volume ratios assessed in our 434

study may be underestimated; for example, Calluna value (7922 m-1) was slightly lower 435

than values reported in other studies (e.g., 10050 m-1 in Fernandes and Rego 1998). This 436

may be because the different methodologies used, and because the scanning procedure 437

can be less accurate than other procedures with direct assessments of particle size. 438

The results presented give a broad view in describing fire danger in litter fuel-beds 439

on the basis of FMC; however, it is worth noting that our results are based on artificial 440

simulations, and further research is needed to contrast our results with real fuel-beds and 441

fires. Previous field studies, however, observed similar moisture of extinction values in 442

litters of maritime pine stands in Portugal (M50 values of 35% to obtain sustaining fires; 443

Fernandes et al., 2008). Davies and Legg (2011) observed significant burning and 444

smouldering of Pleurocarpus mosses at FMC less than 70%; i.e., similar values to our 445

Mmax value of 71.4% observed for Sphagnum in this study. In addition, our M50 values 446

are within the ranges observed in laboratory experiments testing different soil fuel-beds 447

(Lin, 1999; Plucinski and Anderson, 2008). 448

Peat is the deepest strata of the fuel-complex, and it is usually covered by litter and 449

vegetation. It is, therefore unlikely that fires start in this layer directly from small 450

ignition sources such as accidentally-dropped embers or cigarettes ends. In addition, the 451

thresholds of initial flaming ignition and sustained ignition observed for this kind of 452

source was restricted to low FMC values (M50 of 34.9% and 21.6% respectively). 453

However, it is more likely that peat ignition occurs when the litter layer is smouldering; 454

when this occurs wildfire spread will occur upwards into the canopy, then laterally 455

through the canopy and litter-bed and downwards into the peat. The energy available in 456

20

this situation would be expected to be much greater than that provided from small 457

ignition sources (cigarette ends, embers, etc.). Previous studies with more intense 458

ignition sources (greater size and longer duration times: i.e., a coil spiral of 10 mm of 459

diameter, 95 mm long and heated during 30 min) showed that peat was able to ignite at 460

FMC of approximately 115% (Frandsen, 1997; Rein et al., 2008). Therefore, further 461

studies disentangling fire transmission from the litter layer to peat are needed. Fuel-beds 462

with different flammability properties may, therefore, show variable efficiency in fire 463

propagation within British heathlands and moorlands. 464

465

4.2. Fire danger in Calluna vegetation 466

Dead-fuels play a fundamental role in the probability of ignition of Calluna vegetation, 467

being influenced by both FMC and dead-fuel proportion. In fact, it has been proposed 468

that the most likely point where fire starts in the vegetation strata is within these dead 469

fuels (Davies and Legg, 2011). This laboratory approximation determined the values of 470

FMC and dead-fuel proportion that influences these ignition processes. M50 values were 471

variable depending on the source of ignition. When a smouldering source was used, an 472

increasing dead-fuel proportion increased the M50 from 19% to 35%. In contrast, the 473

proportion of dead fuel had little effect when a flaming source was used, where M50 474

remained stable at ca. 30%. These results suggest that management strategies to keep 475

heath/moorlands in a “young state” with less than 20% dead-fuel may be an effective 476

measure for reducing wildfire risk (i.e. building phase – Watt, 1947). Similar 477

management suggestions were proposed for U. europaeus gorse in northern Spain 478

(Marino et al. 2011). 479

Our laboratory experiments used a representative fuel bulk density but clearly 480

variations in this parameter may modulate fire ignition (Marino et al. 2011; Weise et al., 481

21

2005) and further research modeling this effect is needed. Other parameters such as the 482

dead-fuel continuity or the crown base height may also influence fire initiation 483

(Plucinski et al., 2010). Our work, therefore, needs corroboration in field-based studies. 484

However, previous field studies have reported that both fire ignition and sustained 485

spread are correlated strongly with the moisture content of the dead-fuel in the canopy. 486

Davies and Legg (2011) observed in Calluna-dominated ecosystems that fire ignition 487

failed at FMC greater than ca. 70%, but fires started to develop at 60% FMC. These 488

results are in the same order of magnitude as the Mmax observed here (ca. 40-50%), 489

given that the field studies would overestimate FMC because live-fuels in the lower 490

canopy were included. The important role of dead-fuels and their FMC in fire ignition 491

and spread have been also reported for other shrub-dominated systems, for example the 492

Mediterranean gorse (U. parviflorus; Baeza et al., 2002) and the European gorse (Ulex 493

europaeus; Anderson and Anderson, 2010). 494

495

4.3. Effect of the ignition source in fire danger 496

Ignition source is very important in determining the probability of ignition in both 497

peat/litter fuel-beds and vegetation. Smouldering sources were more effective in 498

igniting peat/litter fuel-beds (i.e., igniting them at higher FMC). These sources may be 499

in contact with the soil fuel-bed all along its surface, and therefore, penetrate deeper into 500

the fuel as it is consumed. In contrast, the flaming sources produce a flame plume that is 501

not constantly in intimate contact with the soil fuel-bed, and hence it transfers less 502

energy to the underlying fuel. However, an ignition source can proceed from glowing 503

embers that combine an initial flaming phase with a later smouldering phase (Marino et 504

al. 2010). Further efforts are, therefore, needed to disentangle this interaction. 505

22

For shrubs, the contact of smouldering sources with fuel is restricted to the source 506

surface area, whereas a flame plume can contact vertically with fuel surfaces higher in 507

the vegetation strata, and hence ignite them more efficiently. This is likely to be the 508

reason why ignition in smouldering sources was influenced by the proportion of dead-509

fuel. Higher densities and proportions of dead-fuel may be needed to produce the initial 510

flame and burn at higher FMC. Other factors related to the nature of ignition sources, 511

and not studied here, may also be important in fire ignition, for example, source size, 512

shape and the exposure time to the ignition source (Davies and Legg, 2011; Manzello et 513

al. 2006; Plucinski and Anderson, 2008). 514

515

4.4. Implications for fire danger rating systems 516

Wales and England currently use a fire danger rating system (Meteorological Office 517

Fire Severity Index (MOFSI) (http://www.metoffice.gov.uk/weather/uk/firerisk/; 518

accessed 16 August 2013). Based on the Canadian Wildland Fire Information System 519

(CWFIS), this system consists of a series of basic codes and derived meteorological 520

indices which are used to predict wildfire occurrence (van Wagner, 1987). It has been 521

observed, however, that this system is not well adapted to British moorlands and often 522

fails in its predictions (Davies and Legg, 2008). Only one of its basic codes, the Fine 523

Fuel Moisture Code (FFMC), is able to forecast fire occurrence with acceptable 524

accuracy (Davies and Legg, 2008). FFMC is a numeric rating of the moisture content of 525

litter and other cured fine fuels, and it is computed from data on rainfall, relative 526

humidity, wind speed, and temperature collected over the previous 24 h (van Wagner, 527

1987). Moreover, the moisture content of these fuels can be estimated from FFMC 528

values using simple equations (Aguado et al. 2007; van Wagner, 1987). Therefore, M50 529

and Mmax values presented for the peat/litter fuel-beds in this work can assist in 530

23

improving the predictions of heath/moorland fire danger. For example, better warnings 531

of the critical periods when the different fuel beds drop in fire-prone moisture 532

conditions can be provided by estimating the fuel moisture content through FFMC. 533

Nonetheless, further efforts in calibrating the estimated fuel moisture contents with real 534

field data would be needed. It will also be necessary to take into account the different 535

nature of spring and summer wildfires. In spring the canopy often burns independently 536

from the ground layer because the peat/litter fuel-beds are still wet and frozen. 537

Therefore, further studies to ensure that FFMC is well correlated to the moisture content 538

of dead-fuels are needed to predict this kind of canopy fires. 539

540

5. Conclusions 541

This work helps to disentangle the complex interactions generating wildfire on British 542

heathlands and moorlands. There are four important results reported here. First, there 543

were small differences in moisture thresholds where peat/litter fuel-beds start to ignite 544

into a flame (35-59% FMC), however, the probability of sustained ignitions varied 545

across a wider range (19-55% FMC). Second, we demonstrated that flammability (i.e, 546

ignitability, sustainability, consumability and combustibility) of the peat/litter fuel-beds 547

differs depending on the intrinsic characteristics of species making up the fuel layer. 548

These properties were also influenced strongly by their fuel moisture content. Third, in 549

the upper canopy layer, often composed solely of Calluna, the probability of ignition 550

was influenced both by the proportion of dead fuel accumulated within the vegetation 551

and their FMC. Finally, the source of ignition may play a fundamental role in fire risk 552

assessment, since smouldering sources are more efficient in igniting peat/litter fuel-553

beds, but in the Calluna vegetation layer flaming sources are superior. 554

555

24

Acknowledgements 556

V.M. Santana has been supported by a VAli+d post-doctoral grant awarded by the 557

Generalitat Valenciana. We thank the Heather Trust and the University of Liverpool-558

School of Environmental Sciences’s pump-priming fund for financial support. Juan 559

Hidalgo and Phil Robson helped with the laboratory equipment and field collecting 560

respectively. Gemma Curtis and Jean Routly from the School of Veterinary Sciences 561

kindly helped with the use of the bomb calorimeter. Kath Allen provided valuable 562

comments on the manuscript. 563

564

References 565

Aguado, I., Chuvieco, E., Boren, R., Nieto, H., 2007. Estimation of dead fuel moisture 566

content from meteorological data in Mediterranean areas. Application in fire danger 567

assessment. Int. J. Wildland Fire. 16, 390-397. 568

Albertson, K., Aylen, J., Cavan, G., McMorrow, J., 2009. Forecasting the outbreak of 569

moorland wildfires in the English Peak District. J. Environ. Manage. 90, 2642-2651. 570

Allen, K.A., Harris, M.P.K., Marrs, R.H., 2013. Matrix modelling of prescribed burning 571

in Calluna vulgaris-dominated moorland: short burning rotations minimize carbon 572

loss at increased wildfire frequencies. J. Appl. Ecol. 50, 614-624. 573

Anderson, H.E., 1970. Forest fuel ignitability. Fire Technol. 6, 312-319. 574

Anderson, A.J., Anderson, W.R., 2010. Ignition and fire spread thresholds in gorse 575

(Ulex europaeus). Int. J. Wildland Fire. 19, 589-598. 576

Anon., 2007. The Heather and Grass Burning Code, DEFRA, London. 577

25

Baeza, M.J., De Luis, M., Raventos, J., Escarre, A., 2002. Factors influencing fire 578

behaviour in shrublands of different stand ages and the implications for using 579

prescribed burning to reduce wildfire risk. J. Environ. Manage. 65, 199-208. 580

Chandler, C., Cheney, P., Thomas, P., Trabaud, L., Williams D., 1983. Fire in forestry, 581

John Wiley & Sons, New York. 582

Crawley, M.J., 2012. The R Book, Wiley, Chichester. 583

Davies, G.M., Legg, CJ., 2008. Developing a live fuel moisture model for moorland fire 584

danger rating, In: de las Heras, J., Brebbia, C.A., Viegas, D.X. (Eds.), Forest fires: 585

modeling, monitoring and management of forest fires. WIT Transactions on the 586

Environment, vol 119. WIT Press, Southampton, pp. 225-236. 587

Davies, G.M., Legg, CJ., 2011. Fuel moisture thresholds in the flammability of Calluna 588

vulgaris. Fire Technol. 47, 421-436. 589

Davies, G.M., Legg, C.J., Smith, A.A., MacDonald, A.J., 2009. Rate of spread of fires 590

in Calluna vulgaris-dominated mooorlands. J. Appl. Ecol. 46, 1054-1063. 591

Fernandes, P.M, Rego, F., 1998. A new method to estimate fuel surface area to volume 592

ratio using water immersion. Int. J. Wildland Fire. 8, 121-128. 593

Fernandes, P.M, Botelho, H., Rego, F., Loureiro, C., 2008. Using fuel and weather 594

variables to predict the sustainability of surface fire spread in maritime pine stands. 595

Can. J. Forest Res. 38, 190-201. 596

Fernandes, P.M., Cruz, M.G., 2012. Plant flammability experiments offer limited 597

insight into vegetation-fire dynamics interactions. New Phytol. 194, 606-609. 598

Frandsen, W.H., 1997. Ignition probability of organic soils. Can. J. Forest Res. 27, 599

1471-1477. 600

26

Ganteaume, A., Jappiot, M., Lampin-Maillet, C., Curt, T., Borgniet, L., 2011. Effects of 601

vegetation type and fire regime on flammability of undisturbed litter in Southeastern 602

France. Forest Ecol. Manag. 261, 2223-2231. 603

Gimingham, C.H., 1972. Ecology of Heathlands, Chapman & Hall, London. 604

Harris, M.P.K., Allen, K.A., McAllister, H.A., Eyre, G., Le Duc, M.G., Marrs, R.H., 605

2011. Factors affecting moorland plant communities and component species in 606

relation to prescribed burning. J. Appl. Ecol. 48, 1411-1421. 607

Hosmer, D.W., Lemeshow, S., 2000. Applied Logistic Regression, Wiley Interscience, 608

New York. 609

Krawchuk, M.A., Moritz, M.A., Parisien, M-A., Van Dorn, J., Hayhoe, K., 2009. 610

Global Pyrogeography: the current and future distribution of wildfire. PLoS ONE. 4, 611

e5102. 612

Lin, C.C., 1999. Modelling probability of ignition in Taiwan red pine forests. Taiwan 613

Journal of Forest Science. 14, 339-344. 614

Manzello, S.L., Cleary, T.G., Shields, J.R., Yang, J.C., 2006. Ignition of mulch and 615

grasses by firebrands in wildland-urban interface fires. Int. J. Wildland Fire. 15, 427-616

431. 617

Marino, E., Madrigal, J., Guijarro, M., Hernando, C., Diez, C., Fernandez, C., 2010. 618

Flammability descriptors of fine dead-fuels resulting from two mechanical treatments 619

in shrubland: a comparative laboratory study. Int. J. Wildland Fire. 19, 314-324. 620

Marino, E., Guijarro, M., Hernando, C., Madrigal, J., Diez, C., 2011. Fire hazard after 621

prescribed burning in a gorse shrubland: Implications for fuel management. J. 622

Environ. Manage. 92, 1003-1011. 623

27

Martin, R.E., Gordon, D., Gutierrez, M., Lee, D., Molina, D., Schroeder, R., Sapsis, D., 624

Stephens, S., Chambers, M., 1994. Assessing the flammability of domestic and 625

wildland vegetation, Proceedings of the 12th Conference on Fire and Forest 626

Meteorology, pp. 130-137. 627

McMorrow, J., 2011. Wildfire in the United Kingdom: status and key issues, 628

Proceedings of the second conference on the human dimensions of wildland fire, 629

GTR-NRS-P. Vol. 84. 630

Pausas, J.G., Ribeiro, E., 2013. The global fire-productivity relationship. Global Ecol. 631

Biogeogr. 22, 728-736. 632

Plucinski, M.P., Anderson, W.R., 2008. Laboratory determination of factors influencing 633

successful point ignition in the litter layer of shrubland vegetation. Int. J. Wildland 634

Fire. 17, 628-637. 635

Plucinski, M.P., Anderson, W.R., Bradstock, R.A., Gill, A.M., 2010. The initiation of 636

fire spread in shrublands fuels recreated in the laboratory. Int. J. Wildland Fire. 19, 637

512-520. 638

Rein, G., Cleaver, N., Ashton, C., Pironi, P., Torero, J.L., 2008. The severity of 639

smouldering peat fires and damage to the forest soil. Catena. 74, 304-309. 640

Santana, V.M., Baeza, M.J., Vallejo, V.R., 2011. Fuel structural traits modulating soil 641

temperatures in different species patches of Mediterranean Basin shrublands. Int. J. 642

Wildland Fire. 20, 668-677. 643

Scarf, F.R., Westoby, M., 2006. Leaf litter flammability in some semi-arid Australian 644

woodlands. Funct. Ecol. 20, 745-752. 645

Schmuck, G., San-Miguel-Ayanz, J., Camia, A., Durrant, T., Boca, R., Withmore, C., 646

Liberta, G., Coti, P., Schulte E., 2012. Forest fires in Europe, Middle East and North 647

Africa 2011, European Comission–Joint Research Centre, Luxembourg. 648

28

van Altena, C., van Logtestijn, R.S.P., Cornwell, W.K., Cornelissen, J.H.C., 2012. 649

Species composition and fire; non-additive mixture effects on ground fuel 650

flammability. Front. Plant Sci. 3. 651

Van Wagner, C.E., 1987. Development and structure of the Canadian forest fire weather 652

index system, Canadian Forest Service,Vol. 35, Ottawa. 653

Watt, A.S., 1947. Pattern and processes in the plant community. J. Ecol. 59, 615-622. 654

Weise, D.R., Zhou, X., Sun, L. & Mahalingam, S., 2005. Fire spread in chaparral-‘go or 655

no go’. Int. J. Wildland Fire. 14, 99-106. 656

657

658

659

660

29

Table 1. Parameters used as flammability descriptors for peat/litter fuel-beds in 661 experimental simulations. 662 663 Parameter

descriptor

Variable Units Definition

Ignitability Ignition time (TTI)

Rate of spread (ROS)

s

mm min-1

Time elapsed since the ignition source is

applied until flames appear.

Speed at which the combustion front

propagates.

Sustainability Flaming time (FT)

Elevated temperatures (T300)

s

s

Time of flaming combustion.

Time above 300ºC.

Consumability

Mass loss rate (MLR)

Residual mass fraction (RMF)

mg min-1

%

Speed at which fuel is burnt.

Percentage of fuel remaining after fire

Combustibility Flame height (FH)

Maximum temperature (TMAX)

mm

ºC

Maximum height reached during flaming.

Temperature reached during fuel combustion.

664

665

666

667

668

669

670

671

672

673

30

Table 2. Specific traits of peat/litter fuel-beds derived from British heath/moorlands; 674 mean values ± SD are presented. Letters show significant differences among fuel-beds 675 (One-way ANOVA with Bonferroni pair-wise comparisons). 676 677 Species Fuel-bed

bulk density

(kg m-3)

Particle

specific area

(m2 kg-1)

Particle area to

volume ratio

(m-1)

Mineral

content

(%)

Heat of

combustion

(MJ kg-1)

Calluna 112.9±21.5b 34.4±1.4a 7922±551a 3.2±0.4b 21.2±0.7ab

Vaccinium 42.1±8.1c 39.9±1.3a 7580±1539a 5.9±0.1a 19.9±1.1abc

Empetrum 149.5±24.9b 16.7±1.2b 3592±408b 3±0.1b 22.8±0.4a

Ulex 39.9±8.9c 14.6±1.4b 5619±435ab 1.9±0.2c 21.3±1.1ab

Sphagnum 16.2±5.9c 41.9±5.4a 3658±908b 2.8±0.4b 16.6±0.5c

Peat 288.7±134.5a - - 11.2±3.4a 18.5±1.5bc

F 131.1 350.5 31.1 81.4 15.9

P <0.001 <0.001 <0.001 <0.001 <0.001

n >40 5 5 5 3

678

679

680

681

682

683

684

685

686

687

688

689

690

691

692

693

694

695

696

697

698

699

31

Table 3. GLM models relating flammability properties (a. the probability of initial flaming ignition, and b. the probability of sustained ignition)

of a range of different peat/litter fuel-beds derived from British heath/moorlands in relation to fuel moisture content (FMC).

Species Tests (n) Initial

ignition

M50 Mmax Model parameters Pseudo R2 ROC area

Predictor Estimate SE z-value Odds ratio P

(a) Calluna 43 30 53.6 61.3 Intercept 14.46 6.20 2.33 0.019 0.75 0.97

FMC -0.27 0.11 -2.44 0.76 0.015

Vaccinium 40 23 46.8 51.3 Intercept 8.42 2.79 3.02 0.003 0.66 0.95

FMC -0.18 0.06 -3.09 0.83 0.002

Empetrum 43 38 59.2 62.8 Intercept 11.25 3.93 2.86 0.004 0.57 0.95

FMC -0.19 0.07 -2.74 0.83 0.006

Ulex 40 32 51.4 52.9 Intercept 8.74 2.94 2.97 0.003 0.58 0.96

FMC -0.17 0.06 -2.873 0.84 0.004

Sphagnum 40 17 56.5 80.4 Intercept 4.52 1.58 2.86 0.004 0.34 0.86

FMC -0.08 0.03 -3.12 0.92 0.002

Peat 41 13 34.9 60 Intercept 2.78 0.99 2.79 0.005 0.26 0.93

FMC -0.08 0.02 -3.61 0.93 <0.001

Species Tests (n) Sustained

ignition

M50 Mmax Model parameters Pseudo R2 ROC area

Predictor Estimate SE z-value Odds ratio P

(b) Calluna 43 16 26.9 33.2 Intercept 7.79 2.87 2.77 0.007 0.65 0.96

FMC -0.29 0.11 -2.72 0.75 0.006

Calluna 41 13 15.2 12.8 Intercept 3.61 1.31 2.75 0.005 0.59 0.96

(flame) FMC -0.43 0.15 -2.81 0.65 0.005

Vaccinium 40 13 25.1 51.3 Intercept 2.26 0.95 2.37 0.018 0.38 0.86

FMC -0.09 0.03 -3.17 0.91 0.001

Empetrum 43 10 19.1 36.2 Intercept 2.29 1.11 2.05 0.039 0.34 0.86

FMC -0.12 0.04 -3.09 0.89 0.002

Ulex 40 22 34.5 52.9 Intercept 4.49 1.41 3.18 0.001 0.41 0.91

FMC -0.13 0.04 -3.1 0.88 0.002

Sphagnum 40 17 54.6 71.4 Intercept 7.64 2.66 2.87 0.004 0.52 0.93

FMC -0.14 0.05 -2.98 0.87 0.003

Peat 41 9 21.6 46.1 Intercept 2.81 1.24 2.26 0.024 0.60 0.97

FMC -0.13 0.05 -2.49 0.88 0.013

32

Table 4. A GLM model relating the probability of ignition in simulated Calluna

vegetation to the proportion of dead-fuel moisture content and the type of ignition

source. The intercept of this model was the flaming ignition source; df = 235, a Pseudo

R2 = 0.33 and a ROC area = 0.85.

Predictor Model parameters

Estimate SE z value

Odds

ratio P

Intercept 4.05 0.86 4.68 <0.001

Dead-fuel moisture -0.13 0.01 -7.74 0.87 <0.001

Dead-fuel proportion <-0.00 0.01 -0.03 0.99 0.972

Smouldering -2.51 0.94 -2.65 0.08 0.008

Smouldering x Dead-fuel proportion 0.05 0.02 2.41 1.05 0.015

33

FIGURE CAPTIONS

Fig.1. Flammability descriptors variation of different peat/litter fuel-beds as a function of fuel

moisture content: (a) Ignitability, (b) Sustainability, (c) Consumability, (d) Combustibility.

Results shown correspond to tests using smouldering ignition sources.

Fig. 2. Effect of different proportions of dead-fuel and Fuel Moisture Content at which 50%

of ignitions were successful (M50) and the maximum moisture at which a successful ignition

occurred (Mmax) for Calluna vegetation derived from British moorlands.

34

FIGURE 1

35

FIGURE 2